Political leaning and use of public service media in a fragmented media landscape: longitudinal analysis over 30 years in Sweden
Peter M. Dahlgren
2017-07-18
- 0.1 Setup and import
- 0.2 Swedish Television
- 0.2.1 SVT1
- 0.2.2 SVT1 & leaning
- 0.2.3 SVT1 & political interest
- 0.2.4 SVT1 & extremity
- 0.2.5 SVT2 & leaning
- 0.2.6 SVT2 & political interest
- 0.2.7 Rapport & leaning
- 0.2.8 Rapport & political interest
- 0.2.9 Aktuellt & leaning
- 0.2.10 Aktuellt & political interest
- 0.2.11 Aktuellt/Rapport & leaning
- 0.2.12 Combine measure Aktuellt/Rapport
- 0.2.13 Correlation
- 0.3 Swedish Radio
- 0.4 Public service (PS) use
- 0.5 Internet
- 0.6 Social networking sites (SNS)
- 0.7 Trust
- 0.8 Demography
- 0.9 Political polarization
- 0.10 Affective polarization
- 0.11 Issue positions
- 0.12 Descriptives
- 0.13 Animation
- 0.14 Models
- 0.15 Session info
0.1 Setup and import
knitr::opts_chunk$set(echo = TRUE, fig.width=8, fig.height=4)
knitr::opts_knit$set(root.dir = "C:/Users/Peter/Documents/GU/data/som/")library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(rio)
library(ggplot2)
library(RColorBrewer)
library(tidyverse)## Loading tidyverse: tibble
## Loading tidyverse: tidyr
## Loading tidyverse: readr
## Loading tidyverse: purrr## Conflicts with tidy packages ----------------------------------------------## filter(): dplyr, stats
## lag(): dplyr, statstheme_set(theme_minimal())
# Läs in Super-SOM.
if(file.exists("supersom.RData"))
{
print("Load RData")
load("supersom.rdata")
} else {
print("Import")
system.time(supersom <- import("supersom/SND0905_SuperRSOM_v5.sav"))
print("Saving RData")
save.image(file="supersom.rdata")
}## [1] "Load RData"# Recode rank variable into binary categorical factor.
dummy <- function(x, one=c(1, 2, 3, 4), zero=c(6, 7), labels=c("5 days/week or more", "More seldom"))
{
newx <- NA
newx[x %in% one] <- 1
newx[x %in% zero] <- 0
newx <- factor(newx, levels=c(1, 0), labels=labels)
return(newx)
}
# Recode rank variable into trichotomous categorical factor.
trichotomy <- function(x, two, one, zero, labels=c("Daily", "Weekly", "Never/seldom"))
{
newx <- NA
newx[x %in% two] <- 2
newx[x %in% one] <- 1
newx[x %in% zero] <- 0
newx <- factor(newx, levels=c(2, 1, 0), labels=labels)
return(newx)
}0.1.1 Recode Super-SOM 1986-2015
In general:
- 5 days/week or more is defined as at least 5 days per week.
- More seldom is defined as less than 5 days per week, including never.
# Demography.
supersom$ideology <- supersom$ba10 # Self-reported political left/right placement 1-5.
supersom$ideology[supersom$ideology >= 98] <- NA
supersom$polinterest <- supersom$ca10 %>% recode("1"="4", "2"="3", "3"="2", "4"="1") %>% as.numeric()
supersom$polinterest <- factor(supersom$polinterest, levels=c(1, 2, 3, 4), labels=c("No political interest", "Low", "Medium", "High political interest"))
supersom$polinterest.num <- as.numeric(supersom$polinterest) # Numeric 1-4 political interest.
# Public service.
supersom$svt1 <- supersom$jd20a # SVT1
supersom$svt2 <- supersom$jd20b # SVT2 (1992-2014)
supersom$svt1svt2 <- supersom$jd20c # SVT1/SVT2 (2006-2013)
supersom$rapport <- supersom$jh10d # SVT1 Rapport
supersom$aktuellt <- supersom$jh10c # SVT2 Aktuellt
supersom$aktuelltrapport <- supersom$jh10e # Aktuellt/Rapport
supersom$ekot <- supersom$jh10b # SR Ekot
# Trust SVT.
supersom$trust.svt <- supersom$ka502a %>% recode("1"="3", "2"="3", "3"="2", "4"="1", "5"="1")## Warning: Unreplaced values treated as NA as .x is not compatible. Please
## specify replacements exhaustively or supply .defaultsupersom$trust.svt <- factor(supersom$trust.svt, levels=c(3, 2, 1), labels=c("High trust", "Neither low nor high", "Low trust"))
supersom$trust.sr <- supersom$ka502e %>% recode("1"="3", "2"="3", "3"="2", "4"="1", "5"="1")## Warning: Unreplaced values treated as NA as .x is not compatible. Please
## specify replacements exhaustively or supply .defaultsupersom$trust.sr <- factor(supersom$trust.sr, levels=c(3, 2, 1), labels=c("High trust", "Neither low nor high", "Low trust"))
# Internet, social networking sites.
supersom$internet <- supersom$jb19 # Internet use last 12 months (1995-)
supersom$internetnews <- supersom$jb110c # Internet news use (2005-)
supersom$sns <- supersom$jb110p # Social media use, e.g. Facebook, Twitter (2009-)
# Recode into binary.
supersom$svt1.dich <- dummy(supersom$svt1, one=c(2, 3), zero=c(4, 5, 6, 7))
supersom$svt2.dich <- dummy(supersom$svt2, one=c(2, 3), zero=c(4, 5, 6, 7))
supersom$rapport.dich <- dummy(supersom$rapport, one=c(1, 2), zero=c(3, 4, 5, 6))
supersom$aktuellt.dich <- dummy(supersom$aktuellt, one=c(1, 2), zero=c(3, 4, 5, 6))
supersom$ekot.dich <- dummy(supersom$ekot, one=c(1, 2), zero=c(3, 4, 5, 6))
supersom$sns.dich <- dummy(supersom$sns, one=c(7, 6), zero=c(5, 4, 3, 2, 1))
supersom$internet.dich <- dummy(supersom$internet, one=c(5), zero=c(4,3,2,1), labels=c("Use internet daily", "Use internet more seldom"))
supersom$internetnews.dich <- dummy(supersom$internetnews, one=c(7,6), zero=c(5,4,3,2,1))
# Recode into trichotomoy
supersom$svt1.tri <- trichotomy(supersom$svt1, two=c(2), one=c(3, 4, 5), zero=c(6, 7))
supersom$svt2.tri <- trichotomy(supersom$svt2, two=c(2), one=c(3, 4, 5), zero=c(6, 7))
supersom$rapport.tri <- trichotomy(supersom$rapport, two=c(1), one=c(2, 3, 4), zero=c(5, 6))
supersom$aktuellt.tri <- trichotomy(supersom$aktuellt, two=c(1), one=c(2, 3, 4), zero=c(5, 6))
supersom$ekot.tri <- trichotomy(supersom$ekot, two=c(1), one=c(2, 3, 4), zero=c(5, 6))
supersom$sns.tri <- trichotomy(supersom$sns, two=c(7), one=c(6, 5), zero=c(4, 3, 2, 1))
# Issue positions.
supersom$marry <- dummy(supersom$fc10e, one=c(1, 2, 3), zero=c(4), labels=c("Agree", "Disagree")) # "I would not like immigrant marriying family member."
# Combine Aktuellt/Rapport into single variable.
supersom$aktuelltrapport.dich[supersom$year %in% c(1984:2004)] <- ifelse(supersom$aktuellt[supersom$year %in% c(1984:2004)] <= 5 | supersom$rapport[supersom$year %in% c(1984:2004)] <= 5, 1, 0)
supersom$aktuelltrapport.dich[supersom$year %in% c(2005:2015)] <- ifelse(supersom$aktuelltrapport[supersom$year %in% c(2005:2015)] <= 5, 1, 0)
# Public service use: Using Rapport/Aktuellt/Ekot at least 5 days/week (or not).
supersom <- supersom %>%
mutate(psuse = case_when(is.na(aktuelltrapport.dich) & is.na(ekot.dich) ~ 999,
aktuelltrapport.dich == 1 ~ 1, # Aktuellt/Rapport
ekot.dich == 1 ~ 1, # Ekot
TRUE ~ 0))
supersom$psuse <- factor(supersom$psuse, levels=c(1, 0), labels=c("PS news at least 5 days/week", "PS news more seldom"))
# Daily public service use: Using Rapport/Aktuellt/Ekot daily (or not).
supersom <- supersom %>%
mutate(psuse.daily = case_when(jh10b == 1 ~ 1, # Ekot
jh10c == 1 ~ 1, # Aktuellt
jh10d == 1 ~ 1, # Rapport
jh10e == 1 ~ 1, # Aktuellt/Rapport
is.na(jh10b) & is.na(jh10c) & is.na(jh10d) & is.na(jh10e) ~ 999,
TRUE ~ 0))
supersom$psuse.daily <- factor(supersom$psuse.daily, levels=c(1, 0), labels=c("PS news daily", "PS news more seldom"))
supersom$aktuelltrapport.dich <- factor(supersom$aktuelltrapport.dich, levels=c(1, 0), labels=c("Daily/weekly", "Never/seldom"))
# Lef/right block.
supersom$partyblock[supersom$cb10 %in% c(1, 2)] <- 1 # Left: S+V
supersom$partyblock[supersom$cb10 %in% c(3, 4, 5, 6)] <- 2 # Right: C+L+M+KD
supersom$partyblock[supersom$partyblock > 2] <- NA
supersom$partyblock <- factor(supersom$partyblock, levels=1:2, labels=c("Left", "Right"), ordered=FALSE)
supersom$partyblock.num <- as.numeric(supersom$partyblock %>% recode("1"="-1", "2"="-1", "3"="1", "4"="1", "5"="1", "6"="1"))
# Party.
#supersom$party <- factor(supersom$cb10, levels=c(1:7, 10:13, 30), labels=c("Vänsterpartiet", "Socialdemokraterna", "Centerpartiet", "Liberalerna", "Moderaterna", "Kristdemokraterna", "Miljöpartiet", "Sverigedemokraterna", "Feministiskt initiativ", "Piratpartiet", "Ny demokrati", "Övriga"), ordered=FALSE)
supersom$party <- factor(supersom$cb10, levels=c(1:7, 10:13, 30), labels=c("Left Party (V)", "Social Democrats (S)", "Centre Party (C)", "Liberals (L)", "Moderate Party (M)", "Christian Democrats (KD)", "Green Party (MP)", "Sweden Democrats (SD)", "Feminist Initiative (FI)", "Pirate Party (PP)", "New Democracy (NyD)", "Other"), ordered=FALSE)
supersom$partysymbol <- factor(supersom$cb10, levels=c(1:7, 10:13, 30), labels=c("V", "S", "C", "L", "M", "KD", "MP", "SD", "FI", "PP", "NyD", "Other"), ordered=FALSE)
PartyColor <- unlist(list(V = "#DA291C",
S = "#E8112d",
C = "#009933",
L = "#006AB3",
M = "#52BDEC",
KD = "#000077",
MP = "#83CF39",
SD = "#DDDD00",
FI = "#CD1B68",
PP = "#572B85",
NyD = "#572B85",
Other = "#CCCCCC"), use.names=FALSE)
# Ideological extremism (center = 0, toward left/right = 2)
supersom$polextreme <- supersom$ideology %>% recode("1"="2", "2"="1", "3"="0", "4"="1", "5"="2") %>% as.numeric()
# Strong supporter of party?
supersom$partystrength <- supersom$cb20 %>% recode("1"="2", "2"="1", "3"="0")
supersom$partystrength <- factor(supersom$partystrength, levels=c(0,1,2), labels=c("No party supporter", "Weak party supporter", "Strong party supporter"), ordered=FALSE)
# Political leaning: left/center/right.
supersom$leaning <- supersom$ideology %>% recode("1"="1", "2"="1", "3"="2", "4"="3", "5"="3")
supersom$leaning <- factor(supersom$leaning, levels=c(1,2,3), labels=c("Left", "Center", "Right"), ordered=FALSE)
supersom$leaning.num <- as.numeric(supersom$ideology %>% recode("1"="-2", "2"="-1", "3"="0", "4"="1", "5"="2"))
# Like own party & dislike other parties.
supersom$C <- supersom$cb50a
supersom$M <- supersom$cb50b
supersom$V <- supersom$cb50c
supersom$L <- supersom$cb50d
supersom$S <- supersom$cb50e
supersom$MP <- supersom$cb50f
supersom$KD <- supersom$cb50g
#supersom$Nydemok <- supersom$cb50h
supersom$FI <- supersom$cb50i
supersom$SD <- supersom$cb50m
supersom$likeindex <- colSums(rbind(supersom$C, supersom$M, supersom$V, supersom$L, supersom$S, supersom$MP, supersom$KD, supersom$SD, supersom$FI), na.rm = TRUE)
# Factor sex.
supersom$sex <- factor(supersom$sex, levels=c(1, 2), labels=c("Female", "Male"), ordered=FALSE)
# Convert year to date yyyy-mm-dd.
supersom$date <- as.Date(paste(supersom$year, "-01-01", sep=""))0.1.2 Affect
Calculate feelings towards other parties by subtracting feelings toward out-parties from feelings toward the in-party.
Since there are multiple out-parties, add them all together and average them.
# Add all affects.
supersom$affect <- colSums(rbind(supersom$C, supersom$M, supersom$V, supersom$L, supersom$S, supersom$MP, supersom$KD, supersom$SD, supersom$FI), na.rm = TRUE)
# Subtract in-party from total affect (leaving only out-party affect), and then average the out-party affect.
supersom$affect.outparty <- if_else(supersom$partysymbol == "S", (supersom$affect - supersom$S) / 8, supersom$affect)
supersom$affect.outparty <- if_else(supersom$partysymbol == "M", (supersom$affect - supersom$M) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "V", (supersom$affect - supersom$V) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "MP", (supersom$affect - supersom$MP) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "KD", (supersom$affect - supersom$KD) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "FI", (supersom$affect - supersom$FI) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "C", (supersom$affect - supersom$C) / 8, supersom$affect.outparty)
supersom$affect.outparty <- if_else(supersom$partysymbol == "L", (supersom$affect - supersom$L) / 8, supersom$affect.outparty)
# Make new variable for in-party affect.
supersom$affect.inparty <- if_else(supersom$partysymbol == "S", supersom$S, supersom$affect)
supersom$affect.inparty <- if_else(supersom$partysymbol == "M", supersom$M, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "V", supersom$V, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "MP", supersom$MP, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "KD", supersom$KD, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "FI", supersom$FI, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "C", supersom$C, supersom$affect.inparty)
supersom$affect.inparty <- if_else(supersom$partysymbol == "L", supersom$L, supersom$affect.inparty)0.1.3 Public service index
Create a additive public service index, with the use of Rapport/Aktuellt/Ekot which is then averaged.
Continous scale ranging from 1 (never) to 6 (daily).
# First, combine Aktuellt & Rapport (1986-2005). Use the highest frequency from either variable. In other words, if respondent use either Aktuellt or Rapport daily, then the value is daily.
# (Cumbersome coding because bug in case_when, see https://github.com/tidyverse/dplyr/issues/2927)
supersom$aktuelltrapport_early[supersom$aktuellt == 6 | supersom$aktuellt == 6] <- 6
supersom$aktuelltrapport_early[supersom$aktuellt == 5 | supersom$aktuellt == 5] <- 5
supersom$aktuelltrapport_early[supersom$aktuellt == 4 | supersom$aktuellt == 4] <- 4
supersom$aktuelltrapport_early[supersom$aktuellt == 3 | supersom$aktuellt == 3] <- 3
supersom$aktuelltrapport_early[supersom$aktuellt == 2 | supersom$aktuellt == 2] <- 2
supersom$aktuelltrapport_early[supersom$aktuellt == 1 | supersom$aktuellt == 1] <- 1
# Add to variable 1986-2004.
supersom$aktuelltrapport.combined[supersom$year <= 2004] <- supersom$aktuelltrapport_early[supersom$year < 2004]## Warning in supersom$aktuelltrapport.combined[supersom$year <= 2004] <-
## supersom$aktuelltrapport_early[supersom$year < : number of items to replace
## is not a multiple of replacement length# Use the Aktuellt/Rapport measure from 2005 an onwards.
supersom$aktuelltrapport.combined[supersom$year >= 2005] <- supersom$jh10e[supersom$year >= 2005]
# Add values (Aktuellt and Rapport + Ekot) and divide by 2, thus creating a continous interval scale, from 1-6.
supersom$psindex <- (supersom$aktuelltrapport.combined + supersom$ekot) / 2
# Reverse code index, so that 6=daily, 1=never.
supersom$psindex.rev <- min(supersom$psindex, na.rm=TRUE) + max(supersom$psindex, na.rm=TRUE) - supersom$psindex
# Histogram of index.
ggplot(supersom, aes(psindex.rev)) +
geom_histogram() +
labs(title="Histogram of public service index",
y="Count",
x="Public service index (higher value = more frequent use)")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 15061 rows containing non-finite values (stat_bin).
library(moments)
mean(supersom$psindex.rev, na.rm=TRUE)## [1] 3.690832sd(supersom$psindex.rev, na.rm=TRUE)## [1] 1.38461skewness(supersom$psindex.rev, na.rm=TRUE)## [1] 0.03839314kurtosis(supersom$psindex.rev, na.rm=TRUE)## [1] 2.1249420.2 Swedish Television
0.2.1 SVT1
supersom %>%
select(date, svt1.dich) %>%
na.exclude() %>%
group_by(date, svt1.dich) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(svt1.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom") +
labs(title="Tittar på SVT", x="Year", y="Percent")
0.2.2 SVT1 & leaning
supersom %>%
select(date, svt1.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, svt1.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(svt1.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="SVT1 + leaning",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.2.3 SVT1 & political interest
supersom %>%
select(date, svt1.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, svt1.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(svt1.dich))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="SVT1 & political interest",
x="Year",
y="Percent") +
facet_grid(~ polinterest)
0.2.4 SVT1 & extremity
Are public service viewers more extreme in their views?
supersom %>%
select(date, svt1.dich, polextreme) %>%
na.exclude() %>%
group_by(svt1.dich, date) %>%
summarize(meanextremism = mean(polextreme, na.rm=TRUE)) %>%
ggplot(aes(date, meanextremism, color=svt1.dich)) +
geom_line(alpha=0.5) +
geom_smooth(method="loess") +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "none", axis.text.x = element_text(angle=90)) +
labs(title="SVT1 + ideological extremism",
x="Year",
y="Ideological extremism") +
facet_grid(~ svt1.dich)
0.2.5 SVT2 & leaning
supersom %>%
select(date, svt2.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, svt2.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(svt2.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="SVT2 + leaning",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.2.6 SVT2 & political interest
supersom %>%
select(date, svt2.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, svt2.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(svt2.dich))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="SVT2 + political interest",
x="Year",
y="Percent") +
facet_grid(~ polinterest)
Swedish Television news programmes Rapport and Aktuellt in SVT1 and SVT2, respectively.
0.2.7 Rapport & leaning
supersom %>%
select(date, rapport.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, rapport.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(rapport.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use SVT Rapport",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.2.8 Rapport & political interest
supersom %>%
select(date, rapport.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, rapport.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(rapport.dich))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use SVT Rapport",
x="Year",
y="Percent") +
facet_grid(~ polinterest)
0.2.9 Aktuellt & leaning
supersom %>%
select(date, aktuellt.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, aktuellt.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(aktuellt.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use Aktuellt",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.2.10 Aktuellt & political interest
supersom %>%
select(date, aktuellt.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, aktuellt.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(aktuellt.dich))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use Aktuellt",
x="Year",
y="Percent") +
facet_grid(~ polinterest)
0.2.11 Aktuellt/Rapport & leaning
supersom %>%
select(date, aktuelltrapport.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, aktuelltrapport.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(aktuelltrapport.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Aktuellt/Rapport",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.2.12 Combine measure Aktuellt/Rapport
Combine the two survey questions from 1986-2004 (Aktuellt and Rapport, respectively) with the single survey question from 2005-2014 (Aktuellt/Rapport, combined).
supersom %>%
select(date, aktuelltrapport.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, aktuelltrapport.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(aktuelltrapport.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Aktuell/Rapport",
x="Year",
y="Percent") +
#geom_vline(xintercept = as.numeric(as.Date("2005-01-01")), color="black", alpha=0.9, linetype=2) + # Vertical line that indicate change in survey question
facet_grid(~ leaning)
supersom %>%
select(date, aktuelltrapport.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, aktuelltrapport.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(aktuelltrapport.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Aktuellt/Rapport",
x="Year",
y="Percent") +
geom_vline(xintercept = as.numeric(as.Date("2005-01-01")), color="black", alpha=0.9, linetype=2) +
facet_grid(~ polinterest)
df <- supersom %>% select(polextreme, polinterest.num) %>% na.exclude()
corr.df <- cor.test(df$polextreme, df$polinterest.num)$parameter[[1]]
corr.p <- cor.test(df$polextreme, df$polinterest.num)$p.value
corr.r <- cor(df$polextreme, df$polinterest.num)0.2.13 Correlation
Correlation between political extremity and political interest: r(97193) = 0.2548581, p=0.
0.3 Swedish Radio
0.3.1 Ekot & leaning
supersom %>%
select(date, ekot.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, ekot.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(ekot.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Ekot", x="Year", y="Percent") +
facet_grid(~ leaning)
0.3.2 Ekot & political interest
supersom %>%
select(date, ekot.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, ekot.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(ekot.dich))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90, vjust = 0.5)) +
labs(title="Ekot", x="Year", y="Percent") +
facet_grid(~ polinterest)
0.4 Public service (PS) use
Use any of Aktuellt, Rapport or Ekot.
0.4.1 Frequency
Frequency table of public service news use.
supersom %>%
select(year, psuse) %>%
group_by(year, psuse) %>%
na.exclude() %>%
count() %>%
spread(psuse, n)## # A tibble: 30 x 3
## # Groups: year [30]
## year `PS news at least 5 days/week` `PS news more seldom`
## * <dbl> <int> <int>
## 1 1986 1594 21
## 2 1987 1636 13
## 3 1988 1613 16
## 4 1989 1545 16
## 5 1990 1547 16
## 6 1991 1539 20
## 7 1992 1835 22
## 8 1993 1777 39
## 9 1994 1628 36
## 10 1995 1699 42
## # ... with 20 more rows0.4.2 PS & leaning
supersom %>%
select(date, psuse, leaning) %>%
na.exclude() %>%
group_by(leaning, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
#geom_text(aes(label=round(percent), alpha=0.7), nudge_y=-6, size=2) +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="5 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use of public service news programs",
subtitle=NULL,
x="Year",
y="Percent") +
facet_wrap(~ leaning)
print(paste("n =", supersom %>%
select(date, psuse, leaning) %>%
na.exclude() %>% count()))## [1] "n = 95462"For the latest year, is the difference statistically significant between the left and right?
left <- supersom %>% filter(year == max(year) & leaning == "Left") %>% select(psuse) %>% na.exclude()
center <- supersom %>% filter(year == max(year) & leaning == "Center") %>% select(psuse) %>% na.exclude()
right <- supersom %>% filter(year == max(year) & leaning == "Right") %>% select(psuse) %>% na.exclude()
t.test(x=as.numeric(center$psuse), y=as.numeric(left$psuse), alternative=c("two.sided"), conf.level=0.95)##
## Welch Two Sample t-test
##
## data: as.numeric(center$psuse) and as.numeric(left$psuse)
## t = 4.0702, df = 4216.9, p-value = 4.784e-05
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 0.01665289 0.04760469
## sample estimates:
## mean of x mean of y
## 1.092959 1.060830t.test(x=as.numeric(center$psuse), y=as.numeric(right$psuse), alternative=c("two.sided"), conf.level=0.95)##
## Welch Two Sample t-test
##
## data: as.numeric(center$psuse) and as.numeric(right$psuse)
## t = 5.1399, df = 3946.9, p-value = 2.882e-07
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 0.02374627 0.05303310
## sample estimates:
## mean of x mean of y
## 1.092959 1.054569t.test(x=as.numeric(left$psuse), y=as.numeric(right$psuse), alternative=c("two.sided"), conf.level=0.95)##
## Welch Two Sample t-test
##
## data: as.numeric(left$psuse) and as.numeric(right$psuse)
## t = 0.98434, df = 5090.1, p-value = 0.325
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.006208472 0.018730249
## sample estimates:
## mean of x mean of y
## 1.060830 1.0545690.4.3 PS & political interest
supersom %>%
select(date, psuse, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="5 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use of public service news programs",
subtitle=NULL,
x="Year",
y="Percent") +
facet_grid(~ polinterest)
Frequency table of public service news use 1986 and 2015.
print(paste("n =", supersom %>%
select(date, psuse, polinterest) %>%
na.exclude() %>% count()))## [1] "n = 97257"supersom %>%
select(year, psuse, polinterest) %>%
na.exclude() %>%
group_by(polinterest, year, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
filter(year %in% c(1986, 2015)) %>%
spread(psuse, n)## # A tibble: 16 x 5
## # Groups: polinterest, year [8]
## polinterest year percent `PS news at least 5 days/week`
## * <fctr> <dbl> <dbl> <int>
## 1 No political interest 1986 2.6845638 NA
## 2 No political interest 1986 97.3154362 145
## 3 No political interest 2015 30.2663438 NA
## 4 No political interest 2015 69.7336562 288
## 5 Low 1986 0.8498584 NA
## 6 Low 1986 99.1501416 700
## 7 Low 2015 7.9604131 NA
## 8 Low 2015 92.0395869 2139
## 9 Medium 1986 0.6633499 NA
## 10 Medium 1986 99.3366501 599
## 11 Medium 2015 4.2438893 NA
## 12 Medium 2015 95.7561107 3565
## 13 High political interest 1986 3.4965035 NA
## 14 High political interest 1986 96.5034965 138
## 15 High political interest 2015 5.0886662 NA
## 16 High political interest 2015 94.9113338 1231
## # ... with 1 more variables: `PS news more seldom` <int>0.4.4 PS & extremity
supersom %>%
select(date, psuse, polextreme) %>%
na.exclude() %>%
group_by(psuse, date) %>%
summarize(meanextremism = mean(polextreme, na.rm=TRUE)) %>%
ggplot(aes(date, meanextremism, color=psuse)) +
geom_line(alpha=0.5) +
geom_smooth(method="loess") +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "none", axis.text.x = element_text(angle=90)) +
labs(title="Public service use and ideological extremity",
x="Year",
y="Ideological extremism") +
facet_grid(~ psuse)
print(paste("n =", supersom %>%
select(date, psuse, polextreme) %>%
na.exclude() %>% count()))## [1] "n = 95462"0.4.5 PS & party
supersom %>%
select(date, psuse, party) %>%
na.exclude() %>%
group_by(party, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
#filter(!party %in% c("Feministiskt initiativ", "Piratpartiet", "Ny demokrati", "Övriga")) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="5 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use of public service news programs",
subtitle=NULL,
x="Year",
y="Percent") +
facet_wrap(~ party, ncol = 4)
print(paste("n =", supersom %>%
select(date, psuse, party) %>%
na.exclude() %>%
#filter(!party %in% c("Feministiskt initiativ", "Piratpartiet", "Ny demokrati", "Övriga")) %>%
count()))## [1] "n = 89929"0.4.6 Correlation: SVT trust & Aktuellt/Rapport
# jh10e = Use of Aktuellt/Rapport.
# ka502a = Trust in Swedish Television (SVT).
df <- supersom %>% select(ka502a, jh10e) %>% na.exclude()
corr.df <- cor.test(df$ka502a, df$jh10e)$parameter[[1]]
corr.p <- cor.test(df$ka502a, df$jh10e)$p.value
corr.r <- cor(df$ka502a, df$jh10e)Correlation between trust in SVT and use of Aktuellt/Rapport: r(9177) = 0.2324653, p=7.003867810^{-113}.
0.4.7 Correlation: SR trust & Ekonyheterna
# jh10b = Use of Ekonyheterna
# ka502e = Trust in Swedish Radio (SR).
df <- supersom %>% select(ka502e, jh10b) %>% na.exclude()
corr.df <- cor.test(df$ka502e, df$jh10b)$parameter[[1]]
corr.p <- cor.test(df$ka502e, df$jh10b)$p.value
corr.r <- cor(df$ka502e, df$jh10b)Correlation between trust in SR and use of Ekonyheterna: r(6934) = 0.3077295, p=4.972765810^{-152}.
0.4.8 PS & party support
supersom %>%
select(date, psuse, partystrength) %>%
na.exclude() %>%
group_by(partystrength, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use of public service news programs",
subtitle="Use any of Aktuellt, Rapport or Ekot.",
x="Year",
y="Percent") +
facet_wrap(~ partystrength)
print(paste("n =", supersom %>%
select(date, psuse, partystrength) %>%
na.exclude() %>% count()))## [1] "n = 91375"df <- supersom %>%
select(partystrength, psuse) %>%
na.exclude() %>%
mutate(partystrength = as.numeric(partystrength), psuse=as.numeric(psuse))
corr.df <- cor.test(df$partystrength, df$psuse)$parameter[[1]]
corr.p <- cor.test(df$partystrength, df$psuse)$p.value
corr.r <- cor(df$partystrength, df$psuse)Correlation between public service news use and party identification/support: r(91373) = -0.0367499, p=1.091096210^{-28}.
0.5 Internet
0.5.1 Internet news + leaning
supersom %>%
select(date, internetnews.dich, leaning) %>%
na.exclude() %>%
group_by(leaning, date, internetnews.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(internetnews.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#005800", "#9cfb9c"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use internet news + leaning",
x="Year",
y="Percent") +
facet_grid(~ leaning)
0.5.2 Internet news + party
supersom %>%
select(date, internetnews.dich, party) %>%
na.exclude() %>%
group_by(party, date, internetnews.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(internetnews.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#005800", "#9cfb9c"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use internet news + party",
x="Year",
y="Percent") +
facet_wrap(~ party)
0.5.3 Internet news + political interest
supersom %>%
select(date, internetnews.dich, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, internetnews.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(internetnews.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#005800", "#9cfb9c"), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Use internet news + political interest",
x="Year",
y="Percent") +
facet_grid(~ polinterest)
0.5.4 Internet news + PS use
supersom %>%
select(date, internetnews.dich, psuse) %>%
na.exclude() %>%
group_by(psuse, date, internetnews.dich) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(internetnews.dich))) +
geom_line(size=1.2) +
scale_color_manual(values = c("#005800", "#9cfb9c"), name="Use of internet news") +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Use of internet news among public service (PS) news users",
x="Year",
y="Percent") +
facet_grid(~ psuse)
0.7 Trust
0.7.1 SVT
supersom %>%
select(date, trust.svt) %>%
na.exclude() %>%
group_by(date, trust.svt) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.svt))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.7), nudge_y=6) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "right", axis.text.x = element_text(angle=0)) +
labs(title="Trust SVT", x="Year", y="Percent") 
supersom %>%
select(date, trust.svt, leaning) %>%
na.exclude() %>%
group_by(leaning, date, trust.svt) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.svt))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.7), nudge_y=6) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SVT", x="Year", y="Percent") +
facet_wrap(~leaning)
supersom %>%
select(date, trust.svt, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, trust.svt) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.svt))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.7), nudge_y=6) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SVT", x="Year", y="Percent") +
facet_grid(~polinterest)
supersom %>%
select(date, trust.svt, party) %>%
na.exclude() %>%
group_by(party, date, trust.svt) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.svt))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.7), nudge_y=10) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Förtroende för SVT",
x="Year",
y="Percent",
caption="Data: Super-SOM 1986-2015") +
facet_wrap(~party)## Warning: Removed 7 rows containing missing values (geom_text).
supersom %>%
select(date, trust.svt, sns.dich) %>%
na.exclude() %>%
group_by(sns.dich, date, trust.svt) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.svt))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
#geom_smooth(method="lm", alpha=0.5) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SVT and social networking site use",
x="Year",
y="Percent") +
facet_wrap(~sns.dich)
0.7.2 SR
supersom %>%
select(date, trust.sr) %>%
na.exclude() %>%
group_by(date, trust.sr) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.sr))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "right", axis.text.x = element_text(angle=0)) +
labs(title="Trust SR", x="Year", y="Percent") 
supersom %>%
select(date, trust.sr, leaning) %>%
na.exclude() %>%
group_by(leaning, date, trust.sr) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.sr))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SR", x="Year", y="Percent") +
facet_wrap(~leaning)
supersom %>%
select(date, trust.sr, polinterest) %>%
na.exclude() %>%
group_by(polinterest, date, trust.sr) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.sr))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SR", x="Year", y="Percent") +
facet_grid(~polinterest)
supersom %>%
select(date, trust.sr, party) %>%
na.exclude() %>%
group_by(party, date, trust.sr) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.sr))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Förtroende för SR",
x="Year",
y="Percent",
caption="Data: Super-SOM 1986-2015") +
facet_wrap(~party)## Warning: Removed 3 rows containing missing values (geom_text).
supersom %>%
select(date, trust.sr, sns.dich) %>%
na.exclude() %>%
group_by(sns.dich, date, trust.sr) %>%
summarize(n = n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(date, percent, color=factor(trust.sr))) +
geom_line(size=1, alpha=1) +
geom_point(size=2) +
geom_text(aes(label=round(percent), alpha=0.8), nudge_y=8, size=3.5) +
#geom_smooth(method="lm", alpha=0.5) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle = 0)) +
labs(title="Trust SR, by social networking site use",
x="Year",
y="Percent") +
facet_wrap(~sns.dich)
0.8 Demography
0.8.1 Age
supersom$age3a <- rio::factorize(supersom$age3a)
supersom %>%
select(date, psuse, age8a) %>%
na.exclude() %>%
group_by(age8a, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Ues of public service news programs and age",
subtitle=NULL,
x="Year",
y="Percent") +
facet_wrap(~ age8a)
print(paste("n =", supersom %>%
select(date, psuse, age8a) %>%
na.exclude() %>% count()))## [1] "n = 97830"0.8.2 Sex
supersom$sex <- rio::factorize(supersom$sex)
supersom %>%
select(date, psuse, sex) %>%
na.exclude() %>%
filter(as.numeric(sex) < 3) %>%
group_by(sex, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Ues of public service news programs and female/male",
subtitle="Use any of Aktuellt, Rapport or Ekot.",
x="Year",
y="Percent") +
facet_wrap(~ sex)
print(paste("n =", supersom %>%
select(date, psuse, sex) %>%
na.exclude() %>% count()))## [1] "n = 98565"0.8.3 Education
supersom$edu3 <- rio::factorize(supersom$edu3)
supersom %>%
select(date, psuse, edu3) %>%
na.exclude() %>%
group_by(edu3, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Ues of public service news programs and education",
subtitle="Use any of Aktuellt, Rapport or Ekot.",
x="Year",
y="Percent") +
facet_wrap(~ edu3)
print(paste("n =", supersom %>%
select(date, psuse, edu3) %>%
na.exclude() %>% count()))## [1] "n = 94621"0.8.4 Social class
supersom$occgr <- rio::factorize(supersom$occgr)
supersom %>%
select(date, psuse, occgr) %>%
na.exclude() %>%
group_by(occgr, date, psuse) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
ggplot(aes(date, percent, color=factor(psuse))) +
geom_line(size=1.2) +
#geom_smooth(method="lm") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
scale_y_continuous(labels=scales::comma, limit=c(0, 100), breaks=seq(0, 100, 20)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Ues of public service news programs and social class",
subtitle="Use any of Aktuellt, Rapport or Ekot.",
x="Year",
y="Percent") +
facet_wrap(~ occgr)
print(paste("n =", supersom %>%
select(date, psuse, occgr) %>%
na.exclude() %>% count()))## [1] "n = 85508"0.9 Political polarization
0.9.1 Party support
supersom %>%
select(date, year, partystrength) %>%
group_by(year, partystrength) %>%
na.exclude() %>%
summarize(n=NROW(partystrength)) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(year, percent)) +
geom_line(aes(linetype=partystrength, color=partystrength), size=1.2) +
theme(legend.position = "right") +
labs(title="Party support",
x="Year",
y="Percent")
0.9.2 Leaning
Political (ideological) leaning on the left-right scale (1-5).
supersom %>%
select(date, year, ideology) %>%
group_by(year, ideology) %>%
summarize(n=NROW(ideology)) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(ideology, percent)) +
geom_area(fill="#666666") +
labs(title="Distribution of political ideology by year",
x="Left/right ideology",
y="Percent") +
facet_wrap(~year)## Warning: Removed 31 rows containing missing values (position_stack).
supersom %>%
select(date, year, leaning) %>%
group_by(year, leaning) %>%
na.exclude() %>%
summarize(n=NROW(leaning)) %>%
mutate(percent = (n / sum(n)) * 100) %>%
ggplot(aes(year, percent, color=leaning)) +
geom_line(aes(linetype=leaning), size=1.2) +
theme(legend.position = "bottom") +
labs(title="Political leaning by year",
x="Year",
y="Percent")
We should expect an increase in standard deviation over time if people are becoming more polarized along the left-right scale.
supersom %>%
select(ideology, date) %>%
group_by(date) %>%
summarize(meanideology = mean(ideology, na.rm=TRUE), ideology_sd = sd(ideology, na.rm=TRUE)) %>%
ggplot(aes(date, ideology_sd)) +
geom_line() +
#geom_smooth(method="lm") +
geom_smooth(method="lm", formula=y ~ splines::ns(x, df=2)) + # Use cubic with 2 df.
#geom_smooth(method="loess") +
labs(title="Left/right political polarization",
x="Year",
y="Left-right ideology standard deviation")
0.9.3 Ideological consistency
Ideological consistency, or sorting, where individuals align themselves to parties consistent with their ideology.
supersom %>%
select(year, ideology, partyblock) %>%
#na.exclude() %>%
group_by(year, partyblock, ideology) %>%
summarize(n=n()) %>%
mutate(percent = (n / sum(n)) * 100) %>%
na.exclude() %>%
ggplot() +
geom_area(aes(x=ideology, y=percent, fill=factor(partyblock)), alpha=.8, position="identity") +
scale_color_manual(values=c("#CD614E", "#1B2B5F")) +
scale_fill_manual(values=c("#CD614E", "#1B2B5F")) +
theme(legend.position = "bottom") +
labs(title="Political leaning by party block",
subtitle=NULL,
x="Left/right leaning",
y="Percent",
caption="Left block: V, S. Right block: C, L, KD, M") +
facet_wrap(~ year)
Correlation between leaning and party block identification over time. If polarization is increasing, correlation should increase over time.
df <- supersom %>%
select(leaning.num, partyblock.num, date) %>%
group_by(date, leaning.num, partyblock.num)
df.corr <- plyr::ddply(df, "date", summarise, corr=cor(leaning.num, partyblock.num, method="pearson", use="na.or.complete"))
df.corr %>% mutate(date = format(date, "%Y"))## date corr
## 1 1986 0.7425293
## 2 1987 0.7257491
## 3 1988 0.7373112
## 4 1989 0.6921111
## 5 1990 0.7101502
## 6 1991 0.7518332
## 7 1992 0.7221279
## 8 1993 0.7050289
## 9 1994 0.7409642
## 10 1995 0.7447632
## 11 1996 0.6914982
## 12 1997 0.7020527
## 13 1998 0.7388663
## 14 1999 0.7312772
## 15 2000 0.7117998
## 16 2001 0.7128053
## 17 2002 0.7598317
## 18 2003 0.6852369
## 19 2004 0.7240041
## 20 2005 0.7062038
## 21 2006 0.7686267
## 22 2007 0.7426668
## 23 2008 0.7388060
## 24 2009 0.7537652
## 25 2010 0.7811740
## 26 2011 0.7575875
## 27 2012 0.7622348
## 28 2013 0.7627043
## 29 2014 0.7937577
## 30 2015 0.7765756df.corr %>%
ggplot(aes(date, corr)) +
geom_line() +
#geom_point(size=2) +
geom_smooth(method="lm", formula = y ~ splines::ns(x, df=2)) + # Cubic spline w/ 2 df
#geom_text(aes(label=round(corr, 2)), nudge_y=0) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="3 year") +
theme(legend.position = "right", axis.text.x = element_text(angle=0)) +
labs(title="Ideological consistency / sorting",
subtitle=NULL,
x="Year",
y="Pearson correlation")
Compare high and low public service users.
Too few in the never/seldom category to be meaningful. Therefore, those who use PS news daily is contrasted with those who use it more seldom.
supersom %>%
select(leaning.num, psuse.daily, partyblock.num, date) %>%
group_by(date, psuse.daily, leaning.num) %>%
plyr::ddply(c("date", "psuse.daily"), summarise, corr=cor(leaning.num, partyblock.num, method="pearson", use="na.or.complete")) %>%
na.exclude() %>%
ggplot(aes(date, corr)) +
geom_line() +
geom_smooth(method="lm") +
#geom_point(size=2) +
#geom_text(aes(label=round(corr, 2)), nudge_y=0) +
scale_y_continuous(labels=scales::comma, breaks=seq(0, 1, 0.05)) +
scale_x_date(date_label="%Y", date_breaks="3 year") +
theme(legend.position = "right", axis.text.x = element_text(angle=0)) +
labs(title="Ideological consistency (sorting)",
subtitle=NULL,
x="Year",
y="Pearson Correlation") +
facet_grid(~ psuse.daily)
Compare high and low internet users.
supersom %>%
select(leaning.num, internetnews.dich, partyblock.num, date) %>%
group_by(date, internetnews.dich, leaning.num, partyblock.num) %>%
plyr::ddply(c("date", "internetnews.dich"), summarise, corr=cor(leaning.num, partyblock.num, method="pearson", use="na.or.complete")) %>%
na.exclude() %>%
ggplot(aes(date, corr)) +
geom_line() +
geom_smooth(method="lm") +
#geom_point(size=2) +
#geom_text(aes(label=round(corr, 2)), nudge_y=0) +
scale_y_continuous(labels=scales::comma, breaks=seq(0, 1, 0.02)) +
scale_x_date(date_label="%Y", date_breaks="3 year") +
theme(legend.position = "right", axis.text.x = element_text(angle=0)) +
labs(title="Ideological consistency (sorting)",
subtitle=NULL,
x="Year",
y="Pearson Correlation") +
facet_grid(~ internetnews.dich)
# Get block party + ideology for all years.
df <- supersom %>%
select(date, year, ideology, partyblock) %>%
filter(partyblock == "Left" | partyblock == "Right") %>% # Only include left + right parties, exclude small parties.
group_by(year, partyblock) %>%
summarize(n=NROW(ideology)) %>%
mutate(percent = (n / sum(n)) * 100)
# Sort after year, decreasing.
df <- within(df, year <- ordered(year, levels=rev(sort(unique(year)))))
# Fix order of the geom_bar.
df$position[df$partyblock == "Left"] <- 2## Warning: Unknown or uninitialised column: 'position'.df$position[df$partyblock == "Right"] <- 1
# Stacked bar plot.
ggplot(df, aes(x=factor(year), y=percent, fill=partyblock, group=position)) + # group=position stacks the geom_bar.
geom_bar(stat="identity") +
scale_color_manual(values=c("red", "blue")) +
scale_fill_manual(values=c("red", "blue")) +
#scale_y_continuous(trans="reverse") +
theme(legend.position="bottom", legend.title = element_blank()) +
geom_hline(yintercept=50, color="white", alpha=0.5) +
geom_text(data=df[df$partyblock=="Left", ], y=5, aes(label=formatC(round(percent, 1), format='f', digits=1)), color="white", alpha=.5, position="identity") + # Add percent to bar
geom_text(data=df[df$partyblock=="Right", ], y=95, aes(label=formatC(round(percent, 1), format='f', digits=1)), color="white", alpha=.4, position="identity") + # Add percent to bar
labs(title="Support for party block over time",
x="Year",
y="Percent") +
coord_flip()
# Note: Keep trailing zeros (i.e., 43.0): formatC(round(percent, 1), format='f', digits=1)
# Order by increasing year in facet_wrap.
df$year <- factor(df$year, levels = sort(unique(df$year), decreasing = TRUE))
# Facet by year.
ggplot(df, aes(x=partyblock, y=percent, color=partyblock, fill=partyblock)) +
geom_bar(stat="identity") +
scale_color_manual(values=c("red", "blue")) +
scale_fill_manual(values=c("red", "blue")) +
theme_bw() +
labs(title="Support for party block over time",
x="Left/right ideology",
y="Percent") +
facet_wrap(~year, strip.position = "top")
0.10 Affective polarization
Like and dislike of political parties.
library(reshape2)##
## Attaching package: 'reshape2'## The following object is masked from 'package:tidyr':
##
## smithssupersom %>%
select(C, M, V, L, S, MP, KD, SD, FI, year) %>%
melt(id="year", measure.vars=c("C", "M", "V", "L", "S", "MP", "KD", "SD", "FI")) %>%
filter(value < 10) %>%
ggplot(aes(x=variable, y=value, fill=variable)) +
geom_boxplot() +
theme(legend.position = "none") +
labs(title="Like and dislike of party by year",
x="Party identification",
y="Like/dislike") +
coord_flip() +
facet_wrap(~ year)## Warning: attributes are not identical across measure variables; they will
## be dropped
supersom %>%
select(date, likeindex) %>%
na.exclude() %>%
dplyr::group_by(date) %>%
dplyr::summarize(meanlike = mean(likeindex, na.rm=TRUE)) %>%
ggplot(aes(date, meanlike)) +
geom_line() +
geom_point(size=2) +
#geom_smooth(method="lm", alpha=0.3) +
#scale_color_manual(values = c("#000000", "#CCCCCC"), name="Social networking site use") +
#scale_y_continuous(labels=scales::comma, limits=c(-10, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Like/dislike",
x="Year",
y="Mean value -45 to +45")
supersom %>%
select(date, leaning, likeindex) %>%
na.exclude() %>%
dplyr::group_by(date, leaning) %>%
dplyr::summarize(meanlike = mean(likeindex, na.rm=TRUE)) %>%
ggplot(aes(date, meanlike)) +
geom_line() +
geom_point(size=2) +
#geom_smooth(method="lm", alpha=0.3) +
#scale_color_manual(values = c("#000000", "#CCCCCC"), name="Social networking site use") +
#scale_y_continuous(labels=scales::comma, limits=c(-10, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Like/dislike",
x="Year",
y="Mean value -45 to +45") +
facet_grid(~ leaning)
supersom %>%
select(C, M, V, L, S, MP, KD, SD, FI, year, party) %>%
melt(id="party", measure.vars=c("C", "M", "V", "L", "S", "MP", "KD", "SD", "FI")) %>%
filter(value < 10) %>%
na.exclude() %>%
ggplot(aes(x=variable, y=value, fill=variable)) +
geom_boxplot() +
labs(title="Like and dislike of parties by party", x="Party", y="Like/dislike") +
theme(legend.position = "none") +
coord_flip() +
facet_wrap(~ party)## Warning: attributes are not identical across measure variables; they will
## be dropped
supersom %>%
select(C, M, V, L, S, MP, KD, SD, FI, party, date) %>%
melt(id=c("party", "date"), measure.vars=c("C", "M", "V", "L", "S", "MP", "KD", "SD", "FI")) %>%
filter(value < 10) %>%
na.exclude() %>%
group_by(party, date) %>%
summarize(meanvalue = mean(value, na.rm=TRUE)) %>%
ggplot(aes(date, meanvalue, color=party)) +
geom_line(size=1) +
#geom_point(size=1.2) +
#geom_smooth(method="lm", alpha=0.3) +
#scale_color_manual(values = c("#000000", "#CCCCCC"), name="Social networking site use") +
scale_color_manual(values = unlist(PartyColor, use.names = FALSE)) +
#scale_y_continuous(labels=scales::comma, limits=c(0, 10)) +
#scale_x_date(date_label="%Y", date_breaks="1 year") +
geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
theme(legend.position = "none", axis.text.x = element_text(angle=0)) +
labs(title="Like/dislike of political parties, by each party",
x="Year",
y="Mean value -5 to +5") +
facet_wrap(~ party)## Warning: attributes are not identical across measure variables; they will
## be dropped
0.10.1 Party affect index
# Affect out-party.
supersom %>%
select(date, party, partysymbol, affect.outparty) %>%
filter(partysymbol %in% c("S", "V", "MP", "L", "C", "KD", "M", "SD")) %>%
na.exclude() %>%
group_by(date, party) %>%
summarize(val = mean(affect.outparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="red") +
geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
geom_smooth(method="lm") +
#scale_y_continuous(labels=scales::comma, limit=c(-5, 5)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Out-party affect",
subtitle=NULL,
x="Year",
y="Out-party affect (mean)") +
facet_wrap(~ party, ncol=4)
# Affect in-party.
supersom %>%
select(date, party, partysymbol, affect.inparty) %>%
filter(partysymbol %in% c("S", "V", "MP", "L", "C", "KD", "M", "SD")) %>%
na.exclude() %>%
group_by(date, party) %>%
summarize(val = mean(affect.inparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="green") +
geom_smooth(method="lm") +
geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
#scale_y_continuous(labels=scales::comma, limit=c(-11, 11)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="In-party affect",
subtitle=NULL,
x="Year",
y="In-party affect (mean)") +
facet_wrap(~ party, ncol=4)
# Affect out-party + internet.
supersom %>%
select(date, internet.dich, affect.outparty) %>%
na.exclude() %>%
group_by(date, internet.dich) %>%
summarize(val = mean(affect.outparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="red") +
geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Out-party affect + internet",
subtitle=NULL,
x="Year",
y="Out-party affect (mean)") +
facet_wrap(~ internet.dich, ncol=4)
# Affect out-party + social networking sites.
supersom %>%
select(date, sns.dich, affect.outparty) %>%
na.exclude() %>%
group_by(date, sns.dich) %>%
summarize(val = mean(affect.outparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="red") +
geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Out-party affect + SNS",
subtitle=NULL,
x="Year",
y="Out-party affect (mean)") +
facet_wrap(~ sns.dich, ncol=4)
# Affect in-party + social networking sites.
supersom %>%
select(date, sns.dich, affect.inparty) %>%
na.exclude() %>%
group_by(date, sns.dich) %>%
summarize(val = mean(affect.inparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="darkgreen", alpha=0.2) +
geom_smooth(method="lm", alpha=0.2, color="black") +
#scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="In-party affect + SNS",
subtitle=NULL,
x="Year",
y="In-party affect (mean)") +
facet_wrap(~ sns.dich, ncol=4)
# Affect in-party + PS.
gg1 <- supersom %>%
select(date, psuse.daily, affect.inparty) %>%
na.exclude() %>%
group_by(date, psuse.daily) %>%
summarize(val = mean(affect.inparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="red", alpha=0.2) +
geom_smooth(method="lm", alpha=0.2, color="black") +
#geom_hline(yintercept = 0, alpha=0.4) +
#scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="In-party affect",
subtitle=NULL,
x="Year",
y="Affect (mean)") +
facet_wrap(~ psuse.daily, ncol=4)
# Affect out-party + PS
gg2 <- supersom %>%
select(date, psuse.daily, affect.outparty) %>%
na.exclude() %>%
group_by(date, psuse.daily) %>%
summarize(val = mean(affect.outparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="blue", alpha=0.2) +
geom_smooth(method="lm", alpha=0.2, color="black") +
#geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
#scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="4 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="Out-party affect",
subtitle=NULL,
x="Year",
y="Affect (mean)") +
facet_wrap(~ psuse.daily, ncol=4)
library(gridExtra)##
## Attaching package: 'gridExtra'## The following object is masked from 'package:dplyr':
##
## combinegrid.arrange(gg1, gg2)
# Affect in-party + political interest
gg1 <- supersom %>%
select(date, polinterest, affect.inparty) %>%
na.exclude() %>%
group_by(date, polinterest) %>%
summarize(val = mean(affect.inparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="red", alpha=0.2) +
geom_smooth(method="lm", alpha=0.2, color="black") +
#geom_hline(yintercept = 0, alpha=0.4) +
#scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="8 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="In-party affect",
subtitle=NULL,
x="Year",
y="Affect (mean)") +
facet_grid(~ polinterest)
# Affect out-party + political interest
gg2 <- supersom %>%
select(date, polinterest, affect.outparty) %>%
na.exclude() %>%
group_by(date, polinterest) %>%
summarize(val = mean(affect.outparty)) %>%
ggplot(aes(date, val)) +
geom_line(size=1.2, color="blue", alpha=0.2) +
geom_smooth(method="lm", alpha=0.2, color="black") +
#geom_hline(yintercept = 0, alpha=0.4, linetype=2) +
#scale_y_continuous(labels=scales::comma, limit=c(-2, 2)) +
scale_x_date(date_label="%Y", date_breaks="8 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Out-party affect",
subtitle=NULL,
x="Year",
y="Affect (mean)") +
facet_grid(~ polinterest)
library(gridExtra)
grid.arrange(gg1, gg2)
0.11 Issue positions
Beliefs, views and attitudes on various issues.
0.11.1 “I would not like immigrant marriying family member.”
High and low internet use.
supersom %>%
select(date, marry, internet.dich) %>%
na.exclude() %>%
group_by(internet.dich, date, marry) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
filter(marry == "Agree") %>%
ggplot(aes(date, percent)) +
geom_line(size=1.2, color="#005800") +
#geom_point(size=1.2) +
scale_y_continuous(labels=scales::comma, breaks=seq(0, 100, 5)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
#scale_color_manual(values = "#005800", name=NULL) +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="\"I would not like an immigrant marrying a family member\"",
subtitle=NULL,
x="Year",
y="Percent agreement") +
facet_grid(~ internet.dich)
High and low public service news use.
supersom %>%
select(date, marry, psuse) %>%
na.exclude() %>%
group_by(psuse, date, marry) %>%
summarize(n = n()) %>%
mutate(percent = n / sum(n) * 100) %>%
filter(marry == "Agree") %>%
ggplot(aes(date, percent)) +
geom_line(size=1.2, color="#106FAF") +
#geom_point(size=1.2) +
scale_y_continuous(labels=scales::comma, breaks=seq(0, 100, 5)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
scale_color_manual(values = sort(brewer.pal(3, "Paired")), name=NULL) +
theme(legend.position = "bottom", axis.text.x = element_text(angle=0)) +
labs(title="\"I would not like an immigrant marrying a family member\"",
subtitle=NULL,
x="Year",
y="Percent agreement") +
facet_grid(~ psuse)
supersom %>%
select(date, leaning, fc10e) %>%
na.exclude() %>%
filter(fc10e < 10) %>%
group_by(date, leaning) %>%
summarize(meanvalue = mean(fc10e, na.rm=TRUE)) %>%
ggplot(aes(date, meanvalue)) +
geom_line(size=1) +
geom_point(size=3) +
#geom_smooth(method="lm", alpha=0.3) +
scale_color_manual(values = c("#000000", "#CCCCCC")) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="I would not like immigrant marriying family member",
x="Year",
y="Mean value") +
facet_grid(~ leaning)
0.11.2 “Immigrants should be able to practice their religion freely”
supersom %>%
select(date, leaning, fc10b) %>%
na.exclude() %>%
filter(fc10b < 10) %>%
group_by(date, leaning) %>%
summarize(meanvalue = mean(fc10b, na.rm=TRUE)) %>%
ggplot(aes(date, meanvalue)) +
geom_line(size=1) +
geom_point(size=3) +
#geom_smooth(method="lm", alpha=0.3) +
scale_color_manual(values = c("#000000", "#CCCCCC")) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Immigrants should be able to practice their religion freely",
x="Year",
y="Mean value") +
facet_grid(~ leaning)
0.11.3 “Internet essential for news”
supersom %>%
select(date, leaning, bb100k) %>%
na.exclude() %>%
filter(bb100k < 50) %>%
group_by(date, leaning) %>%
summarize(meaninternet = mean(bb100k, na.rm=TRUE)) %>%
ggplot(aes(date, meaninternet)) +
geom_line(size=1) +
geom_point(size=3) +
#geom_smooth(method="lm", alpha=0.3) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name="Social networking site use") +
scale_y_continuous(labels=scales::comma, limits=c(0, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Internet is essential to stay informed about news",
x="Year",
y="Mean value (1-10)") +
facet_grid(~ leaning)
0.11.4 “Internet can only complement news papers, radio and television”
supersom %>%
select(date, leaning, bb100h) %>%
na.exclude() %>%
filter(bb100h < 50) %>%
group_by(date, leaning) %>%
summarize(meanvalue = mean(bb100h, na.rm=TRUE)) %>%
ggplot(aes(date, meanvalue)) +
geom_line(size=1) +
geom_point(size=3) +
#geom_smooth(method="lm", alpha=0.3) +
scale_color_manual(values = c("#000000", "#CCCCCC"), name="Social networking site use") +
scale_y_continuous(labels=scales::comma, limits=c(0, 10)) +
scale_x_date(date_label="%Y", date_breaks="1 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Internet can only complement news papers, radio and television",
x="Year",
y="Mean value (1-10)") +
facet_grid(~ leaning)
0.12 Descriptives
Demography, summary statistics, number of respondents etc.
0.12.1 Political interest
supersom %>%
group_by(date) %>%
select(date, polinterest.num) %>%
summarize(meaninterest=mean(polinterest.num, na.rm=TRUE)) %>%
ggplot(aes(date, meaninterest)) +
geom_line(alpha=0.4) +
geom_smooth(method = "loess", level=0.95, color="blue", alpha=0.3) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
geom_vline(xintercept = as.numeric(as.Date("2004-01-01")), color="black", alpha=0.9, linetype=2) +
#geom_vline(xintercept = as.numeric(as.Date("2006-01-01")), color="black", alpha=0.9, linetype=4)
annotate("text", x=as.Date("2008-06-01"), y=2.35, label="Facebook launches 2004") +
labs(title="Mean political interest",
subtitle="Loess smoothing with 95 % CI.",
x="Year",
y="Political interest (1–4)")
supersom %>%
group_by(leaning, date) %>%
select(date, polinterest.num) %>%
na.exclude() %>%
summarize(meaninterest=mean(polinterest.num, na.rm=TRUE)) %>%
ggplot(aes(date, meaninterest)) +
geom_line(alpha=0.4) +
#geom_smooth(method = "loess", level=0.95, color="blue", alpha=0.3) +
#geom_smooth(method = "lm", level=0.95, color="blue", alpha=0.3) +
geom_smooth(method = "lm", formula=y ~ splines::ns(x, df=2), level=0.95, color="blue", alpha=0.3) +
scale_y_continuous(labels=scales::comma, limits = c(2, 3)) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(axis.text.x = element_text(angle=90)) +
labs(title="Political interest (mean)",
subtitle=NULL,
x="Year",
y="Political interest (1–4)") +
#geom_vline(xintercept = as.numeric(as.Date("2004-01-01")), color="black", alpha=0.9, linetype=2) +
#geom_vline(xintercept = as.numeric(as.Date("2006-01-01")), color="black", alpha=0.9, linetype=4)
#annotate("text", x=as.Date("2008-06-01"), y=2.35, label="Facebook launches 2004") +
facet_wrap(~ leaning)## Adding missing grouping variables: `leaning`
0.12.2 Ideology
supersom %>%
group_by(date) %>%
select(date, ideology) %>%
summarize(meanideology=mean(ideology, na.rm=TRUE)) %>%
ggplot(aes(date, meanideology)) +
geom_line() +
#geom_smooth(method = "lm", level=0.95) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
geom_hline(yintercept=3, linetype=2) +
labs(title="Ideology per year (mean)", x="Year", y="Ideology (1=left, 5=right)")
0.12.2.1 Extreme ideology
Ideological extremism by year.
supersom %>%
group_by(date) %>%
select(date, polextreme) %>%
summarize(meanextremism=mean(polextreme, na.rm=TRUE)) %>%
ggplot(aes(date, meanextremism)) +
geom_line(alpha=0.4) +
geom_smooth(method = "loess", level=0.95, color="blue", alpha=0.3) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
labs(title="Ideological extremism (mean)",
subtitle=NULL,
x="Year",
y="Ideological extremism (0-2)") +
geom_vline(xintercept = as.numeric(as.Date("2004-01-01")), color="black", alpha=0.9, linetype=2) +
#geom_vline(xintercept = as.numeric(as.Date("2006-01-01")), color="black", alpha=0.9, linetype=4)
annotate("text", x=as.Date("2008-06-01"), y=1.23, label="Facebook launches 2004")
Ideological extremism by party block.
supersom %>%
group_by(leaning, date) %>%
select(date, polextreme, leaning) %>%
na.exclude() %>%
summarize(meanextremism=mean(polextreme, na.rm=TRUE)) %>%
filter(meanextremism > 0) %>%
ggplot(aes(date, meanextremism, color=leaning)) +
geom_line(alpha=0.8) +
geom_smooth(method = "loess", level=0.95, alpha=0.15) +
scale_y_continuous(labels=scales::comma) +
scale_x_date(date_label="%Y", date_breaks="2 year") +
theme(legend.position = "bottom", axis.text.x = element_text(angle=90)) +
labs(title="Ideological extremism (mean)",
subtitle=NULL,
x="Year",
y="Ideological extremism (0–2)") +
facet_grid(~leaning)
0.12.3 Respondents
supersom %>%
ggplot(aes(year)) +
geom_bar(position="stack", aes(fill=factor(formular))) +
theme(legend.position = "none") +
labs(title="Respondents per year and survey",
subtitle="Colors indicate the different surveys.",
x="Year",
y="Respondents")
Sample:
print(paste("Total n =", nrow(supersom)))## [1] "Total n = 103589"print(paste("Age mean =", mean(supersom$age, na.rm=TRUE)))## [1] "Age mean = 48.5126053726528"print(paste("Age SD =", sd(supersom$age, na.rm=TRUE)))## [1] "Age SD = 18.1446076498253"0.12.4 Response rate
Data from page 542 table 5: http://som.gu.se/digitalAssets/1487/1487725_531-560-frida-vernersdotter--metod.pdf
surveys <- data.frame(
year=1986:2015,
date=as.Date(seq.Date(as.Date("1986-01-01"), as.Date("2015-01-01"), by = "year")),
responserate=c(65,67,66,63,63,63,67,66,61,63,63,63,64,63,59,61,60,61,60,58,55,57,54,55,56,52,52,49,51,51)
)
print(paste("Response rate M =", mean(surveys$responserate)))## [1] "Response rate M = 59.6"print(paste("Response rate SD =", sd(surveys$responserate)))## [1] "Response rate SD = 5.24963052887107"surveys %>% ggplot(aes(date, responserate)) +
geom_line() +
scale_y_continuous(limits = c(0, 100)) +
labs(title="Response rate", x="Year", y="Percent")
0.13 Animation
(Not included)
0.14 Models
0.14.1 Functions
# Append significance stars depending on p-value.
sigstars <- function(p) {
p <- as.numeric(p)
if(p < 0.001) { return ("***") }
if(p < 0.01) { return ("**") }
if(p < 0.05) { return ("*") }
return("")
}
# Coefficient plot for LM.
coefplot <- function(model, intercept=FALSE, sort=FALSE, title="Predictors", caption="*p<0.05 **p<0.01 ***p<0.001. Bars represent 95 % C.I.", labels=NULL, nudge_y=0.3, nudge_x=0) {
library(broom)
if(!is.null(labels)) {
names(model$coefficients) <- labels # Rename labels.
}
df <- tidy(model, conf.int=TRUE)
if(intercept == FALSE){
df <- df %>% filter(term != "(Intercept)")
}
df <- df %>% mutate(label = paste(round(estimate, 2), sapply(p.value, FUN=sigstars), sep=""))
if(sort == TRUE) {
df$term <- reorder(df$term, df$estimate) # Sort by estimate, descdending order.
}
ggplot(df, aes(estimate, term)) +
geom_point(size=3, fill="black") +
geom_errorbarh(aes(xmin=conf.low, xmax=conf.high), width=0.2, height=0, lwd=1, lty=1, col="black") +
geom_vline(xintercept = 0, color="black", alpha=0.5, linetype=2) +
geom_text(aes(label=label, alpha=1), nudge_y=nudge_y, nudge_x=nudge_x) +
theme(legend.position = "none") +
labs(title=title,
caption=caption,
x="Regression coefficient (unstandardized)",
y=NULL)
}
# Coefficient plot for MLM.
coefplot_mlm <- function(model, intercept=FALSE, sort=FALSE, title="Predictors", caption="Bars represent 95 % C.I.", labels=NULL, nudge_y=0.3, nudge_x=0) {
library(broom)
if(!is.null(labels)) {
names(model$coefficients) <- labels # Rename labels.
}
df <- tidy(model, conf.int=TRUE)
if(intercept == FALSE){
df <- df %>% filter(term != "(Intercept)")
}
df <- df %>% mutate(label = paste(round(estimate, 2), sep=""))
if(sort == TRUE) {
df$term <- reorder(df$term, df$estimate) # Sort by estimate, descdending order.
}
df %>%
na.exclude() %>%
ggplot(aes(estimate, term)) +
geom_point(size=3, fill="black") +
geom_errorbarh(aes(xmin=conf.low, xmax=conf.high), width=0.2, height=0, lwd=1, lty=1, col="black") +
geom_vline(xintercept = 0, color="black", alpha=0.5, linetype=2) +
geom_text(aes(label=label, alpha=1), nudge_y=nudge_y, nudge_x=nudge_x) +
theme(legend.position = "none") +
labs(title=title,
caption=caption,
x="Estimate",
y=NULL)
}
# Return confidence interval for the PLM's.
# From http://rforpublichealth.blogspot.se/2014/10/easy-clustered-standard-errors-in-r.html
plm_confint <- function(model, vcovCL){
t <- qt(0.975, model$df.residual)
ct <- coeftest(model, vcovCL)
est <- cbind(ct[,1], ct[,1] - t * ct[,2], ct[,1] + t * ct[,2])
colnames(est) <- c("Estimate","LowerCI","UpperCI")
return(est)
}Prepare data for models.
# Set reference categories.
supersom$polinterest <- relevel(supersom$polinterest, ref="No political interest")
supersom$party <- relevel(supersom$party, ref="Social Democrats (S)")
supersom$leaning <- relevel(supersom$leaning, ref="Center")
# Center year 1986 as 0.
supersom$year_centered <- supersom$year - 19860.14.2 Model 1. Pooled cross-section OLS
Simple model with only the main predictors.
# Linear model.
m1 <- lm(psindex.rev ~ leaning + polinterest.num + year_centered, data=supersom)
summary(m1)##
## Call:
## lm(formula = psindex.rev ~ leaning + polinterest.num + year_centered,
## data = supersom)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.5515 -0.9243 0.0441 1.0441 3.2740
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.3115165 0.0181655 127.248 < 2e-16 ***
## leaningLeft 0.0036642 0.0114493 0.320 0.749
## leaningRight 0.0730934 0.0112547 6.494 8.38e-11 ***
## polinterest.num 0.5564094 0.0057784 96.291 < 2e-16 ***
## year_centered -0.0048949 0.0005463 -8.960 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.305 on 85212 degrees of freedom
## (18372 observations deleted due to missingness)
## Multiple R-squared: 0.1054, Adjusted R-squared: 0.1054
## F-statistic: 2511 on 4 and 85212 DF, p-value: < 2.2e-16coefplot(m1, title=NULL, labels=c("(Intercept)", "Leaning left", "Leaning right", "Political interest", "Time (0-30)"))## Warning: Ignoring unknown parameters: width
0.14.3 Model 2
More predictors. Clustered standard errors.
m2 <- lm(psindex.rev ~ factor(polinterest) + sex + factor(age4a) + edu3 + leaning + year_centered + I(year_centered * as.integer(polinterest)), data=supersom)
library(stargazer)##
## Please cite as:## Hlavac, Marek (2015). stargazer: Well-Formatted Regression and Summary Statistics Tables.## R package version 5.2. http://CRAN.R-project.org/package=stargazerstargazer(m2, type="text")##
## ==========================================================================================
## Dependent variable:
## -----------------------------
## psindex.rev
## ------------------------------------------------------------------------------------------
## factor(polinterest)Low 0.355***
## (0.021)
##
## factor(polinterest)Medium 0.626***
## (0.030)
##
## factor(polinterest)High political interest 0.749***
## (0.043)
##
## sexMale 0.146***
## (0.008)
##
## factor(age4a)2 0.632***
## (0.012)
##
## factor(age4a)3 1.205***
## (0.013)
##
## factor(age4a)4 1.751***
## (0.014)
##
## edu3Medel (allt över grundskola men ej högskola/universitet) 0.090***
## (0.012)
##
## edu3Hög (studier/examen från högskola/universitet) 0.158***
## (0.013)
##
## leaningLeft 0.054***
## (0.011)
##
## leaningRight 0.070***
## (0.010)
##
## year_centered -0.037***
## (0.002)
##
## I(year_centered * as.integer(polinterest)) 0.008***
## (0.001)
##
## Constant 2.426***
## (0.029)
##
## ------------------------------------------------------------------------------------------
## Observations 81,484
## R2 0.277
## Adjusted R2 0.277
## Residual Std. Error 1.173 (df = 81470)
## F Statistic 2,396.730*** (df = 13; 81470)
## ==========================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01print(paste("n =", nrow(model.frame(m2))))## [1] "n = 81484"coefplot(m2, sort=TRUE, intercept=TRUE, title=NULL, caption=NULL,
labels=c("(Intercept)", "Low political interest", "Medium political interest", "High political interest", "Male", "30–49 years", "50–64 years", "65 years or older", "Medium education", "High education", "Leaning left", "Leaning right", "Time (0–30)", "Time × Political interest"),
nudge_y=0, nudge_x=0.2)## Warning: Ignoring unknown parameters: width
0.14.4 Model 3 + Party
Add party.
m3 <- lm(psindex.rev ~ factor(polinterest) + sex + factor(age4a) + edu3 + leaning + as.numeric(partystrength) + year_centered + I(year_centered * as.integer(polinterest)) + factor(party), data=supersom)
stargazer(m3, type="text")##
## ==========================================================================================
## Dependent variable:
## -----------------------------
## psindex.rev
## ------------------------------------------------------------------------------------------
## factor(polinterest)Low 0.349***
## (0.023)
##
## factor(polinterest)Medium 0.616***
## (0.032)
##
## factor(polinterest)High political interest 0.749***
## (0.046)
##
## sexMale 0.158***
## (0.009)
##
## factor(age4a)2 0.611***
## (0.012)
##
## factor(age4a)3 1.179***
## (0.013)
##
## factor(age4a)4 1.713***
## (0.015)
##
## edu3Medel (allt över grundskola men ej högskola/universitet) 0.093***
## (0.012)
##
## edu3Hög (studier/examen från högskola/universitet) 0.159***
## (0.014)
##
## leaningLeft 0.063***
## (0.013)
##
## leaningRight 0.046***
## (0.013)
##
## as.numeric(partystrength) -0.001
## (0.006)
##
## year_centered -0.034***
## (0.002)
##
## I(year_centered * as.integer(polinterest)) 0.008***
## (0.001)
##
## factor(party)Left Party (V) -0.044**
## (0.017)
##
## factor(party)Centre Party (C) 0.167***
## (0.021)
##
## factor(party)Liberals (L) 0.087***
## (0.019)
##
## factor(party)Moderate Party (M) -0.015
## (0.017)
##
## factor(party)Christian Democrats (KD) 0.051**
## (0.022)
##
## factor(party)Green Party (MP) -0.088***
## (0.017)
##
## factor(party)Sweden Democrats (SD) -0.206***
## (0.025)
##
## factor(party)Feminist Initiative (FI) -0.144**
## (0.057)
##
## factor(party)Pirate Party (PP) -0.594***
## (0.071)
##
## factor(party)New Democracy (NyD) -0.008
## (0.067)
##
## factor(party)Other -0.163***
## (0.034)
##
## Constant 2.420***
## (0.034)
##
## ------------------------------------------------------------------------------------------
## Observations 74,626
## R2 0.275
## Adjusted R2 0.275
## Residual Std. Error 1.167 (df = 74600)
## F Statistic 1,134.419*** (df = 25; 74600)
## ==========================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01print(paste("n =", nrow(model.frame(m3))))## [1] "n = 74626"coefplot(m3, sort=TRUE, intercept=TRUE, title=NULL, caption=NULL,
labels=c("(Intercept)", "Low political interest", "Medium political interest", "High political interest", "Male", "30–49 years", "50–64 years", "65 years or older", "Medium education", "High education", "Leaning left", "Leaning right", "Party ID strength (1–3)", "Time (0–30)", "Time × Political interest", "Left Party (V)", "Centre Party(C)", "Liberals (L)", "Moderate Party (M)", "Christian Democrats (KD)", "Green Party (MP)", "Sweden Democrats (SD)", "Feminist Initiative (FI)", "Pirate Party (PP)", "New Democracy (NyD)", "Other party"),
nudge_y=0, nudge_x=0.3)## Warning: Ignoring unknown parameters: width
0.14.5 Model 4 + Internet
Add internet news use + social networking site use + SVT trust + SR trust.
m4 <- lm(psindex.rev ~ factor(polinterest) + sex + factor(age4a) + internetnews + sns + edu3 + leaning + as.numeric(partystrength) + factor(party) + factor(trust.svt) + (trust.sr) + year_centered + I(year_centered * as.integer(polinterest)), data=supersom)
stargazer(m4, type="text")##
## ========================================================================================
## Dependent variable:
## ---------------------------
## psindex.rev
## ----------------------------------------------------------------------------------------
## factor(polinterest)Low 0.112
## (0.332)
##
## factor(polinterest)Medium 0.522
## (0.647)
##
## factor(polinterest)High political interest 0.741
## (0.970)
##
## sexMale 0.081**
## (0.034)
##
## factor(age4a)2 0.751***
## (0.050)
##
## factor(age4a)3 1.419***
## (0.054)
##
## factor(age4a)4 1.947***
## (0.063)
##
## internetnews 0.020**
## (0.009)
##
## sns -0.041***
## (0.007)
##
## edu3Medel (allt över grundskola men ej högskola/universitet) 0.039
## (0.056)
##
## edu3Hög (studier/examen från högskola/universitet) 0.028
## (0.060)
##
## leaningLeft -0.131**
## (0.052)
##
## leaningRight 0.019
## (0.049)
##
## as.numeric(partystrength) 0.015
## (0.024)
##
## factor(party)Left Party (V) 0.008
## (0.072)
##
## factor(party)Centre Party (C) 0.019
## (0.084)
##
## factor(party)Liberals (L) -0.027
## (0.079)
##
## factor(party)Moderate Party (M) -0.131**
## (0.062)
##
## factor(party)Christian Democrats (KD) -0.140
## (0.098)
##
## factor(party)Green Party (MP) -0.080
## (0.059)
##
## factor(party)Sweden Democrats (SD) -0.113
## (0.078)
##
## factor(party)Feminist Initiative (FI) -0.134
## (0.166)
##
## factor(party)Pirate Party (PP) -0.502**
## (0.216)
##
## factor(party)Other -0.248
## (0.180)
##
## factor(trust.svt)Neither low nor high -0.103
## (0.063)
##
## factor(trust.svt)Low trust -0.050
## (0.128)
##
## trust.srNeither low nor high -0.428***
## (0.062)
##
## trust.srLow trust -0.651***
## (0.126)
##
## year_centered -0.032
## (0.036)
##
## I(year_centered * as.integer(polinterest)) 0.002
## (0.012)
##
## Constant 3.171***
## (0.632)
##
## ----------------------------------------------------------------------------------------
## Observations 4,876
## R2 0.363
## Adjusted R2 0.359
## Residual Std. Error 1.105 (df = 4845)
## F Statistic 91.944*** (df = 30; 4845)
## ========================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01coefplot(m4, sort=TRUE, title="Predictor of public service news use", nudge_y=0, nudge_x=0.5)## Warning: Ignoring unknown parameters: width
0.14.6 PLM Model
library(plm)## Loading required package: Formula##
## Attaching package: 'plm'## The following objects are masked from 'package:dplyr':
##
## between, lag, leadlibrary(lmtest)## Loading required package: zoo##
## Attaching package: 'zoo'## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numericlibrary(multiwayvcov)
p.df <- supersom %>% select(year, psindex.rev, polinterest.num, year, year_centered) %>% pdata.frame(index=c("year"))
plm1 <- plm(psindex.rev ~ polinterest.num + year_centered, data = p.df)
# OLS coefficients and standard errors clustered by year, from http://www.richard-bluhm.com/clustered-ses-in-r-and-stata-2/
# Compute Stata like df-adjustment.
G <- length(unique(p.df$year))
N <- length(p.df$year)
dfa <- (G / (G - 1)) * (N - 1) / plm1$df.residual
# display with cluster VCE and df-adjustment.
time_c_vcov <- dfa * vcovHC(m3, type = "HC0", cluster = "time", adjust = T)
plm2 <- coeftest(plm1, vcov = time_c_vcov)
coefplot(plm1, title="Predictors of public service use, normal SE's")## Warning: Ignoring unknown parameters: width
#coefplot(plm2, title="Predictors of public service use, clustered SE's")0.14.7 MLM Model 1
Empty model, only dependent variable and year as a random intercept.
library(lme4)## Loading required package: Matrix##
## Attaching package: 'Matrix'## The following object is masked from 'package:tidyr':
##
## expand##
## Attaching package: 'lme4'## The following object is masked _by_ '.GlobalEnv':
##
## dummy## The following object is masked from 'package:rio':
##
## factorizemlm1 <- lmer(psindex.rev ~ 1 + (1 | year), data=supersom, REML=FALSE)
summary(mlm1)## Linear mixed model fit by maximum likelihood ['lmerMod']
## Formula: psindex.rev ~ 1 + (1 | year)
## Data: supersom
##
## AIC BIC logLik deviance df.resid
## 308755.9 308784.1 -154375.0 308749.9 88525
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.0249 -0.8419 -0.1162 0.6406 1.7927
##
## Random effects:
## Groups Name Variance Std.Dev.
## year (Intercept) 0.003972 0.06303
## Residual 1.913791 1.38340
## Number of obs: 88528, groups: year, 30
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 3.69399 0.01262 292.70.14.8 MLM Model 2
Year as random intercept + political interest as fixed slope.
library(lme4)
mlm2 <- lmer(psindex.rev ~ 1 + polinterest + (1 | year), data=supersom, REML=FALSE)
summary(mlm2)## Linear mixed model fit by maximum likelihood ['lmerMod']
## Formula: psindex.rev ~ 1 + polinterest + (1 | year)
## Data: supersom
##
## AIC BIC logLik deviance df.resid
## 294007.5 294063.8 -146997.8 293995.5 87302
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.69499 -0.72829 -0.00125 0.76691 2.72783
##
## Random effects:
## Groups Name Variance Std.Dev.
## year (Intercept) 0.006472 0.08045
## Residual 1.696626 1.30255
## Number of obs: 87308, groups: year, 30
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2.61148 0.02160 120.92
## polinterestLow 0.77277 0.01730 44.68
## polinterestMedium 1.37644 0.01707 80.64
## polinterestHigh political interest 1.73461 0.02018 85.97
##
## Correlation of Fixed Effects:
## (Intr) plntrL plntrM
## polintrstLw -0.659
## polntrstMdm -0.667 0.833
## plntrstHgpi -0.563 0.705 0.7170.14.9 MLM Model 3
Year as random intercept + everything else as fixed slope.
mlm3 <- lmer(psindex.rev ~ 1 + factor(polinterest) + sex + factor(age4a) + edu3 + leaning + as.numeric(partystrength) + I(year_centered * as.numeric(polinterest)) + (1 | year_centered), data=supersom, REML=FALSE)
summary(mlm3)## Linear mixed model fit by maximum likelihood ['lmerMod']
## Formula: psindex.rev ~ 1 + factor(polinterest) + sex + factor(age4a) +
## edu3 + leaning + as.numeric(partystrength) + I(year_centered *
## as.numeric(polinterest)) + (1 | year_centered)
## Data: supersom
##
## AIC BIC logLik deviance df.resid
## 243776.5 243924.6 -121872.3 243744.5 77297
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5534 -0.7189 -0.0678 0.7583 3.6496
##
## Random effects:
## Groups Name Variance Std.Dev.
## year_centered (Intercept) 0.08187 0.2861
## Residual 1.36748 1.1694
## Number of obs: 77313, groups: year_centered, 29
##
## Fixed effects:
## Estimate
## (Intercept) 1.8692775
## factor(polinterest)Low 0.3793572
## factor(polinterest)Medium 0.6618708
## factor(polinterest)High political interest 0.7975130
## sexMale 0.1461993
## factor(age4a)2 0.6247171
## factor(age4a)3 1.1963702
## factor(age4a)4 1.7432904
## edu3Medel (allt över grundskola men ej högskola/universitet) 0.0923657
## edu3Hög (studier/examen från högskola/universitet) 0.1637141
## leaningLeft 0.0535141
## leaningRight 0.0685691
## as.numeric(partystrength) -0.0023585
## I(year_centered * as.numeric(polinterest)) 0.0072388
## Std. Error
## (Intercept) 0.0580260
## factor(polinterest)Low 0.0218768
## factor(polinterest)Medium 0.0309370
## factor(polinterest)High political interest 0.0440827
## sexMale 0.0085950
## factor(age4a)2 0.0121292
## factor(age4a)3 0.0131602
## factor(age4a)4 0.0146536
## edu3Medel (allt över grundskola men ej högskola/universitet) 0.0120914
## edu3Hög (studier/examen från högskola/universitet) 0.0133623
## leaningLeft 0.0111063
## leaningRight 0.0109778
## as.numeric(partystrength) 0.0062511
## I(year_centered * as.numeric(polinterest)) 0.0006621
## t value
## (Intercept) 32.21
## factor(polinterest)Low 17.34
## factor(polinterest)Medium 21.39
## factor(polinterest)High political interest 18.09
## sexMale 17.01
## factor(age4a)2 51.51
## factor(age4a)3 90.91
## factor(age4a)4 118.97
## edu3Medel (allt över grundskola men ej högskola/universitet) 7.64
## edu3Hög (studier/examen från högskola/universitet) 12.25
## leaningLeft 4.82
## leaningRight 6.25
## as.numeric(partystrength) -0.38
## I(year_centered * as.numeric(polinterest)) 10.93##
## Correlation matrix not shown by default, as p = 14 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itcoefplot_mlm(mlm3, title="Fixed effects", caption="Year as random effect", sort=TRUE, nudge_y=0, nudge_x=0.2)## Warning: Ignoring unknown parameters: width
Correlation matrix.
print(mlm3, correlation=TRUE)## Linear mixed model fit by maximum likelihood ['lmerMod']
## Formula: psindex.rev ~ 1 + factor(polinterest) + sex + factor(age4a) +
## edu3 + leaning + as.numeric(partystrength) + I(year_centered *
## as.numeric(polinterest)) + (1 | year_centered)
## Data: supersom
## AIC BIC logLik deviance df.resid
## 243776.5 243924.6 -121872.3 243744.5 77297
## Random effects:
## Groups Name Std.Dev.
## year_centered (Intercept) 0.2861
## Residual 1.1694
## Number of obs: 77313, groups: year_centered, 29
## Fixed Effects:
## (Intercept)
## 1.869277
## factor(polinterest)Low
## 0.379357
## factor(polinterest)Medium
## 0.661871
## factor(polinterest)High political interest
## 0.797513
## sexMale
## 0.146199
## factor(age4a)2
## 0.624717
## factor(age4a)3
## 1.196370
## factor(age4a)4
## 1.743290
## edu3Medel (allt över grundskola men ej högskola/universitet)
## 0.092366
## edu3Hög (studier/examen från högskola/universitet)
## 0.163714
## leaningLeft
## 0.053514
## leaningRight
## 0.068569
## as.numeric(partystrength)
## -0.002359
## I(year_centered * as.numeric(polinterest))
## 0.0072390.15 Session info
Show locale and package versions for reproducibility.
sessionInfo()## R version 3.4.1 (2017-06-30)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 10 x64 (build 15063)
##
## Matrix products: default
##
## locale:
## [1] LC_COLLATE=Swedish_Sweden.1252 LC_CTYPE=Swedish_Sweden.1252
## [3] LC_MONETARY=Swedish_Sweden.1252 LC_NUMERIC=C
## [5] LC_TIME=Swedish_Sweden.1252
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] lme4_1.1-13 Matrix_1.2-10 multiwayvcov_1.2.3
## [4] lmtest_0.9-35 zoo_1.8-0 plm_1.6-5
## [7] Formula_1.2-2 stargazer_5.2 broom_0.4.2
## [10] gridExtra_2.2.1 reshape2_1.4.2 lubridate_1.6.0
## [13] moments_0.14 bindrcpp_0.2 purrr_0.2.2.2
## [16] readr_1.1.1 tidyr_0.6.3 tibble_1.3.3
## [19] tidyverse_1.1.1 RColorBrewer_1.1-2 ggplot2_2.2.1
## [22] rio_0.5.5 dplyr_0.7.1
##
## loaded via a namespace (and not attached):
## [1] httr_1.2.1 jsonlite_1.5 splines_3.4.1
## [4] modelr_0.1.0 assertthat_0.2.0 cellranger_1.1.0
## [7] yaml_2.1.14 backports_1.1.0 lattice_0.20-35
## [10] quantreg_5.33 glue_1.1.1 digest_0.6.12
## [13] rvest_0.3.2 minqa_1.2.4 colorspace_1.3-2
## [16] sandwich_2.3-4 htmltools_0.3.6 plyr_1.8.4
## [19] psych_1.7.5 pkgconfig_2.0.1 SparseM_1.77
## [22] haven_1.1.0 scales_0.4.1 openxlsx_4.0.17
## [25] MatrixModels_0.4-1 mgcv_1.8-17 car_2.1-5
## [28] nnet_7.3-12 lazyeval_0.2.0 pbkrtest_0.4-7
## [31] mnormt_1.5-5 magrittr_1.5 readxl_1.0.0
## [34] evaluate_0.10.1 nlme_3.1-131 MASS_7.3-47
## [37] forcats_0.2.0 xml2_1.1.1 foreign_0.8-69
## [40] tools_3.4.1 data.table_1.10.4 hms_0.3
## [43] stringr_1.2.0 munsell_0.4.3 compiler_3.4.1
## [46] rlang_0.1.1 grid_3.4.1 nloptr_1.0.4
## [49] labeling_0.3 rmarkdown_1.6 boot_1.3-19
## [52] gtable_0.2.0 curl_2.7 R6_2.2.2
## [55] knitr_1.16 bdsmatrix_1.3-2 bindr_0.1
## [58] rprojroot_1.2 stringi_1.1.5 parallel_3.4.1
## [61] Rcpp_0.12.11
0.6 Social networking sites (SNS)
0.6.1 SNS & Extremity
Are social media users more extreme in their views?
0.6.2 SNS t-test
0.6.3 SNS & leaning
Note that the survey question changed in 2014.
0.6.4 SNS & party support
Note that the survey question changed in 2014.
0.6.5 SNS & political interest
Note that the survey question changed in 2014.
0.6.6 SNS & PS use
0.6.7 SNS + internet news
0.6.8 Twitter med
#aktuellttrendTwitter trends collected every hour since 2014.
Same thing again, but relative to all trends that week.